To understand "Deep Learning+", we must go back to the history of deep learning technology's development and adoption in China. In 2006, Hinton and others unexpectedly discovered the entirely new possibilities brought about by multi-layered neural networks, thus propelling machine learning technology, which had already emerged in the 1980s, into a new stage of deep learning. The outstanding performance of deep learning technology in a series of AI testing tasks also rekindled people's expectations for AI. Objectively speaking, the core element for the resurgence of AI technology after experiencing two periods of stagnation is the emergence of deep learning technology.
For many years afterward, deep learning remained a niche field, even within academia. In 2010, Wang Haifeng, who had just joined Baidu, discovered the immense potential of this technology. Baidu quickly recognized his findings and began exploring deep learning technology and applications as early as 2012. In January 2013, Baidu established the world's first dedicated Deep Learning Research Institute (IDL). Simultaneously, Baidu began developing its own deep learning framework. Thus, fortunately and inevitably, thanks to Wang Haifeng's insightful discovery and Baidu's decisive action, Chinese companies did not fall behind in the AI wave and even became global leaders in AI exploration.
The years that followed can be considered the formative period for deep learning technology in China. During this time, the AI industry, represented by Baidu, achieved a series of successes, such as the emergence and development of the independent deep learning framework PaddlePaddle, and the advancements in various deep learning algorithms. With the arrival of the global AI boom in 2017, China released the "New Generation Artificial Intelligence Development Plan," and the strategic importance and social demand for deep learning technology rose accordingly. Facing the new situation, Baidu maximized its accumulated deep learning technology and platform advantages, rapidly carrying out a series of strategic upgrades and industrial growth.
Will AI cause mass unemployment?
I believe that if people react with aversion and rejection to the development of AI, it's only a matter of time before it gradually replaces some people's jobs. However, if we can accept the development of AI, actively understand and use it, and make it a powerful assistant in our daily work and life, then we won't be replaced by AI. Instead, we will be able to create better content with its help.
This is not a helpless situation of "joining when you can't win," but rather an inevitable result of human development.
Just as firearms eventually replaced cold weapons, informationized armies evolved from mechanized armies, and the internet has to some extent replaced traditional media, the reason we are who we are today is also a result of embracing many new things. Moreover, at present, the development of some technologies is at a bottleneck, or their future is blocked by a thin sheet of paper.
For example, computer graphics in the VR field also needs AI to tackle it from another angle. Even John Carmack, a leading figure in computer graphics, is exploring the path of general artificial intelligence, stating that he "wants to try some areas where nobody knows where it will lead."
Whether it's the brutal competition in the chip industry, the AI algorithm race, the knowledge graph contest, or even the reckless spending without a clear direction, the expectations and anxieties brought by AI, and all of humanity's current state, are because no one can predict when AI technology will explode and pierce through the thin veil covering the future.
With the advent of the Internet of Things (IoT) era, MCUs have been given the higher mission of serving as the computing hub on the edge. Whether it's the functionality of the device itself or its role as part of the IoT, IoT devices can no longer function without MCUs in terms of connectivity, interaction, and security.
Now that intelligence meets edge computing, MCUs have taken on a new mission.
With the further improvement of MCU computing power, the main frequency of high-frequency MCUs has been increased to the GHz level, which can now meet the low computing power artificial intelligence needs of edge devices. Integrating artificial intelligence onto MCUs and using only a single chip for edge deployment is becoming a new trend.
In the past few years, many MCU manufacturers, including Renesas, have been actively exploring the combination of MCUs and artificial intelligence.
At the recent ADI MCU Media Workshop, Xin Yi, a senior engineer at ADI China Technical Support Center, introduced ADI's latest edge AI solution, the MAX78000, which embodies this approach.
Xin Yi explained that the MAX78000 integrates a convolutional neural network (CNN) engine, which can meet the needs of edge artificial intelligence applications, and the design principle of the MAX78000 is to minimize the power consumption of the CNN engine.
To achieve this goal, the device employs a dual-core architecture of an Arm Cortex-M4F processor and a 32-bit RISC-V processor, and incorporates a CNN engine.
Xin Yi vividly compared this architecture to "father and mother": the two microcontroller cores are "the mother who buys groceries," while the CNN accelerator is "the father who cooks," working together to complete the computing work of edge intelligence.
